Power trends: Silicon carbide makes the drop

Power trends: Silicon carbide makes the drop

Interviews |
By Nick Flaherty

Based on technology developed at Rutgers University in the US, the new devices are drop in replacements for existing silicon MOSFETs, offering more efficiency and power capability while using the same gate drivers.

This required a whole new approach, and he is realistic about the time it takes for development. “It’s generally in 15 to 20 years for a new material system to demonstrate efficacy to really mainstream,” said Dries (above).  “Diodes were important in the 2000s but the SiC transistor technologies hadn’t taken off and we wanted to do it in a very different way. The first thing we did was go fabless, compared to Cree, Infineon and Rohm.”

“That was particularly effective back then that the wafer diameters matched legacy silicon fabs that were transitioning out, so we started on a 4in GaAs and InP line then onto an automotive qualified 6in line and we have brought both of those up at a time when the demand is surging and what’s driving that is automotive. So much of our business is growing in Asia while i

n the US people are slower to latch onto it. Europe designs in SiC in a big way.”

“To be successful in a fragmented market like this we have to do something unique,” he said. “What we make is not a traditional SiC MOSFET – we make a normally on SiC JFET and co-package with a custom low voltage silicon MOSFET in an always off device. The retained charge (Qrr) is about 3x lower than a silicon MOSFET and so has the gate drive – the downside is the added packaging complexity, but we can use an 8in mainstream silicon foundry for that.”

“What’s happening in the market is the user doesn’t care [what the device structure is]. This lends itself to more sophisticated levels of integration – I can co-package a JFET with on-chip silicon MOSFET with its own gate driver and controller IC, which allows you to put an LLC stage in a package so what we see in the next two or three years is why not have a whole flyback converter in a chip. We are solving all the problems of co-packing already so the next logical level is the integration.”

The other element is modules. “It will be diodes and FETs for the next few years but we are working with costumers on modules in a stack configuration to make very compact devices,” he said. “The silicon MOSFET can be soldered onto the JFET so you eliminate the need for the freewheeling diode in the module. We are not going to market with these modules in the next 18 months but we are working with customers on the first demonstrations of these modules.”

A United SiC module

The first products were diodes for power factor correction (PFC) in high end server power supplies and these are just reaching volume production, but the industry is looking at new designs in renewable energy, telecom power supplies and automotive, first in onboard chargers in the 3kW to  10kW range but also for the traction drives displacing IGBTs.

“For us our largest design cycles have been two years and that’s happening more rapidly because of designs in Asia,” he said. “We have a drop in replacement for superjunction MOSFETs or IGBTs with the same gate drive rather than a redesign. People literally take out the MOSFET or IGBT and instantly get higher efficiency across the entire load range. For example I had a server company in Taiwan take out two 650V MOSFETs in parallel and replace with one 650V device, and we have done this while remaining compatible with competing SiC MOSFETs.”

The topology that has the most traction is totem pole power factor correction, he says. “This was started by the GaN guys at 650V but most customers are not comfortable with the robustness of GaN yet. We are the only company in the SiC that guarantees a short circuit of 4ms with 650V class FETs and we do that through the ACQ101 and go beyond it to drive our FETS into avalanche with 1m times and 1000 hours of continuous running in inverter rectifier loops to go above and beyond.”

He believes the SiC industry will stay on its current 4in and 6in wafers for a considerable time despite the increasing demand for volume production.

The structure of the 3D printed United SiC module 

“Our fab partner has two domestic foundries in US at 4in and 6in,” he said. “The way that bulk silicon is grown through sublimation means the requirements on flatness and warpage are very stringent and typically the stress in a wafer that large makes it almost impossible to grow epi on it in the short term,” he said. “So until that’ s resolved we will be at 6in for a long time and that allows us to leverage low cost 6in fabs around the world.”

The silicon replacement approach means he is also looking at lower voltages.

“One of the time things we are looking at its moving the voltage down rather than up,” he said. “What we have done is drive our die so small I think we can be at price parity with silicon superjunction devices in a couple of years and that drives up the total market by at least a billion dollars. In contrast to the gaN guys the reality is that the die is 6 to 7x larger than our FETs. I’m very bullish on our ability in the 650V higher power sockets. For more consumer oriented designs like chargers, no we won’t be competitive there but we will for anything in the kW class.”

“If you look at the growth curves they have always been over optimistic but right now I think these predictions are behind the curve and in the next couple of years we will pass the $1bn market onto several billion in the early 2020s. We’re in the part of the cycle where it’s all doing in the right direction – we see tremendous growth and opportunity in China and that will drive down the cost.”

But he is also addressing the more traditional high voltage area to open up new solid state applications.

“The bulk of the business is at 1200V but what’s really exciting is using the same JFET for a supercascode. We can source epitaxy up to 1700V. Beyond that the epi becomes much more difficult and 3300V is the commercial limit but we have a unique way around it through the supercascode. You can extend the voltage by adding 1200V normally on JFETs in series in a passive network and that’s a beautiful way of producing 6500V and up discretely and in modules so it’s a cost effective way to access high voltage.”

“We’re engineers and we love solving difficult problems. So the place I see it being used is in solid state power transformers for dynamic power routing.  We have a customer in Germany, Behlke, that is developing very high voltage switches using our devices.”

Related stories:


If you enjoyed this article, you will like the following ones: don't miss them by subscribing to :    eeNews on Google News


Linked Articles